New Machine Learning Model Helps Decode James Webb Telescope Images
A machine learning model named Morpheus is helping scientists classify every pixel of high-resolution images captured by the James Webb Space Telescope. The tool identifies stars and galaxies by adapting computer vision algorithms for astronomical data.
A machine learning model named Morpheus plays a key role in analyzing high-resolution images from NASA’s James Webb Space Telescope. The model acts as a convolutional neural network that classifies every single pixel of these full-color images to identify astronomical objects like stars and galaxies.
Research scientist Ryan Hausen originally developed Morpheus in 2019 by modifying a semantic segmentation algorithm used for Hubble Space Telescope data. The upgraded version, known as Big Morpheus, gains its advanced capabilities through scaling and acceleration on a NVIDIA GPU-enabled supercomputer called lux at the University of California, Santa Cruz.
Beyond space exploration, scientists adapt Morpheus for other scientific domains, such as classifying sea ice when telescopes point back at Earth. Hausen highlights that this type of scalable semantic segmentation offers exciting opportunities for computer scientists to apply machine learning to novel data that ultimately benefits all of humanity.